Toward a Predictive Model of Success in Contingency Management: A Proof of Concept Study Utilizing Behavioral Economic, Clinical Severity, and Alcohol Use Severity Measures

Abstract As contingency management (CM) moves from research to practice, researchers have a responsibility to outline the minimum procedural necessities that lead to an effective, sustainable treatment that can be implemented as a mainstream therapy for substance use disorders. To begin identifying the minimum requirements, the purpose of the current study was to provide framework and a first step toward building a risk calculator that predicts treatment outcomes in CM, and can predict the optimal incentive size to prescribe by evaluating behavioral economic factors, demographic variables, and use severity measures in individuals who completed CM treatment for alcohol use. Participants were 38 individuals enrolled in the active treatment arms of two parent CM studies for reducing alcohol use (Koffarnus et al., 2018; Koffarnus et al., 2021). Participants were 42 years old on average, 55% male, and a majority were white, non-Hispanic. Fifteen candidate predictor variables were assessed for inclusion in the predictive model including demographic variables, use severity scores, and behavioral economic parameters. A logistic regression framework was used to identify top predictive models. Accuracy was assessed by computing receiver operating characteristic (ROC) curves and area under the curves. A model including the delay discounting parameter, log 10 (ED50) of alcohol, participant age, and Beck’s Anxiety Inventory score was predictive of treatment outcomes in the current sample. The results demonstrate the utility of the ROC analysis as a method for identifying a predictive model. Further research is needed to replicate and verify the findings of the current analysis.